When should you start doing paid ads?

Is $5,000 MRR really the point where you should start paid ads?

For a bootstrapped SaaS, around $5,000 MRR is a sensible point to begin serious paid-ad testing, although small experiments can start much earlier.

There is no magic revenue threshold hidden inside Google or Meta. The usefulness of $5,000 MRR comes from something more practical: paid acquisition currently costs enough that a proper test can consume a meaningful chunk of a tiny company's revenue.

WordStream's 2026 Google Ads benchmark, based on more than 13,000 search campaigns, puts the average search cost per click at $5.42 and the average cost per lead at $66.69. Business services, a closer comparison for many SaaS products, reached $93.69 per lead. Search CPCs are now more than twice their 2016 level of $2.32, even though average cost per lead has finally fallen after five consecutive years of increases.

At those prices, $200 of ad spend tells us very little about customer acquisition. Even $1,000 may produce only ten or fifteen business-service leads before we have looked at how many start a trial, pay, renew or churn.

So we would separate early experimentation from real acquisition. A founder can spend $50 before earning a dollar if there is a precise question to answer. Spending several thousand dollars every month deserves a much higher bar.

SaaS MRR	What paid ads make sense for
Below $1,000	Small demand, keyword and message experiments
$1,000–$5,000	Narrow tests around very high-intent audiences
$5,000–$10,000	Serious CAC and payback testing
$10,000–$20,000	Scaling channels that have already worked
$20,000+	MRR itself should no longer decide whether ads scale

Can you start paid ads before $1,000 MRR or even before launching?

Yes, a SaaS can use paid ads before $1,000 MRR, and sometimes before the product exists, when we are deliberately paying to learn whether real demand exists.

The danger starts when clicks and signups are mistaken for proof that the business works.

A recent founder-reported example on Indie Hackers shows the difference nicely. Pregnalyze spent $313 on Google Ads for a consumer product priced at $5.99 for its basic paid offer. The campaign generated 454 clicks and roughly 200 email leads. Only one person paid.

From an advertising dashboard, 200 leads from $313 could look encouraging. From a business perspective, one $5.99 transaction from $313 of spend gives a completely different answer.

Another founder documented testing a SaaS landing page with only €20 of Google Ads before building the product. That is a perfectly reasonable use of advertising. A €20 experiment can tell us whether particular searches exist, whether a message earns clicks and whether anyone leaves an email address. Nobody should extrapolate a scalable CAC from such a tiny sample.

We would happily run experiments like these at $0 MRR. We just would not call them an acquisition engine yet.

How much retention do you need before buying more SaaS customers?

Paid ads become worth scaling once at least one SaaS customer cohort clearly sticks around; otherwise we are paying to fill a bucket that still leaks too quickly.

The current SaaS retention data shows how common that problem is at the beginning.

ChartMogul's dataset of more than 2,500 SaaS companies puts median monthly customer churn at 6.5% for companies below $300,000 ARR. That falls to 3.7% between $1 million and $3 million ARR and 3.1% above $8 million ARR.

A 6.5% monthly churn rate means that, if the rate stayed constant, only about 45% of the starting customers would still be around twelve months later. That gives a young SaaS much less time to recover its advertising costs.

A universal rule such as "never advertise above 3% monthly churn" would create fake precision. Products have different prices, margins and customer lifetimes. What we want to see is much simpler: several cohorts should behave in roughly the same way, a meaningful group of customers should keep paying, and the gross profit from those customers should comfortably cover what we spend to acquire them.

Until we can see that, adding traffic mostly makes the retention problem larger.

How much money do you actually need to test paid ads properly?

For Google Search, roughly $1,500 to $3,000 is a more realistic serious testing budget than a few hundred dollars if we want enough leads to start judging acquisition economics.

The sample-size math is pretty unforgiving.

At WordStream's current $66.69 average Google Search cost per lead, 20 leads cost about $1,334 and 30 cost about $2,001. Using the $93.69 business-services benchmark, those same samples rise to roughly $1,874 and $2,811.

Meta can generate leads more cheaply. WordStream's latest broad Meta lead-campaign benchmark found a $27.66 average cost per lead, although lead quality and purchase intent can be very different from search.

And these are still leads. If 30 leads eventually create five customers, we are estimating CAC from five purchases. One unusually good or bad customer already changes the result a lot.

That is why very small paid campaigns are useful for learning but weak evidence for scaling.

Benchmark	Approx. cost for 20 leads	Approx. cost for 30 leads
Meta lead campaigns at $27.66 CPL	$553	$830
Google Search at $66.69 CPL	$1,334	$2,001
Google business services at $93.69 CPL	$1,874	$2,811

Is $5,000 MRR enough to take paid ads seriously?

Yes, $5,000 MRR is enough for a bootstrapped SaaS to start taking paid acquisition seriously, but we would still treat the first campaigns as tests rather than a permanent growth budget.

Look at the size of the experiment relative to the company.

Thirty leads at the current Google business-services benchmark cost about $2,811. For a company doing $5,000 MRR, that equals 56% of one month's recurring revenue. Even twenty leads cost about $1,874, or 37% of monthly revenue.

Of course, the campaign can run over several months and MRR is revenue rather than available cash. Still, the order of magnitude tells us why $5,000 feels very different from $500 MRR.

The company can finally buy enough traffic to learn something useful without one test automatically becoming absurd relative to the size of the business. By then, it also tends to have more customers from which to estimate churn, conversion and average revenue.

We would concentrate the spend heavily at this stage. If people already search for exactly the problem the SaaS solves, one narrow Google Search campaign is more useful than dividing $2,000 between Google, Meta, LinkedIn, Reddit and five different audiences.

At $5,000 MRR, finding one channel that works is enough.

Why does $10,000 MRR feel much safer for paid ads?

Around $10,000 MRR, paid ads become much easier to test repeatedly because the company usually has both more cash coming in and more customer history behind its numbers.

The same $2,000 experiment that consumes 40% of a $5,000 month consumes 20% of a $10,000 month. More importantly, we are probably making decisions from dozens or hundreds of customers rather than a handful.

That second part can matter even more than the money.

A SaaS doing $10,000 MRR from ten enterprise customers still has very little statistical evidence about retention. Another SaaS could reach the same MRR with 500 customers paying $20 each and understand its cancellation curve extremely well. Two companies can have identical MRR and completely different readiness for ads.

Revenue also does not make mediocre advertising magically good. Angel Match provides a useful recent example. Founder Rashid Khasanov said the SaaS was doing about $37,300 MRR after having run Meta ads for most of 2025 at an average reported ROAS of roughly 1.6x. The company still paused Meta for about a month to examine how profitable those ads really were before resuming them.

That is a company above $30,000 MRR questioning its paid economics. A $10,000 threshold should therefore be treated as financial breathing room, rather than permission to scale whatever campaign happens to be running.

How fast should paid ads pay back their CAC?

For a bootstrapped SMB SaaS, we would want paid CAC back within roughly six months before scaling aggressively; six to twelve months can still be workable when retention and cash flow are strong.

CAC payback simply asks how many months of gross profit it takes to recover what we spent acquiring a customer.

Bessemer's cloud benchmarks classify a CAC payback below six months as best, six to twelve months as better and twelve to eighteen months as good. Its broader cloud guidance recommends staying below roughly twelve months for SMB products, while mid-market and enterprise companies can support longer periods because contracts tend to be larger and customers tend to stay longer.

Paddle independently places a typical SaaS CAC payback around five to twelve months.

The market has also become less forgiving. Benchmarkit's 2025 study of private B2B SaaS companies found that the median company was spending $2 of sales and marketing expense to generate $1 of new-customer ARR, up 14% from the previous year. That includes much more than advertising, but it shows how expensive new revenue has become.

For a small bootstrapped company, waiting twelve or eighteen months to get acquisition cash back can be uncomfortable even if the theoretical lifetime value eventually looks attractive. Six months gives us much more room to keep reinvesting.

Assume an 80% gross margin. A customer paying $20 a month contributes roughly $16 of monthly gross profit, so a six-month payback supports about $96 of CAC. A $500 customer contributes roughly $400 and can support around $2,400 under the same rule.

The difference is enormous even though both companies might have exactly the same total MRR.

Monthly revenue per customer	Monthly gross profit at 80% margin	CAC for 6-month payback	CAC for 12-month payback
$20	$16	$96	$192
$50	$40	$240	$480
$100	$80	$480	$960
$500	$400	$2,400	$4,800
$1,000	$800	$4,800	$9,600

Can high-ticket B2B SaaS start paid ads much earlier?

Yes, high-ticket B2B SaaS can justify paid ads at surprisingly low MRR because a single customer may repay several months of advertising.

A company charging $20,000 a year has a completely different acquisition budget from a $15-per-month app, even if both are currently doing $3,000 MRR.

The freshest large B2B advertising dataset also shows how expensive this game can become. Metadata's newly released benchmark analyzed $57.6 million of 2025 ad spend from 153 B2B advertisers. The typical advertiser paid $202 for a LinkedIn lead. Google Ads reached $524 per lead in the B2B dataset.

Once Metadata followed the campaigns into CRM data, the numbers became much harsher. Across the advertisers with enough revenue data to measure, the reported ad-generated customer cost was $58,887, and each advertising dollar produced only $0.56 of traceable first-year closed-won revenue. LinkedIn alone showed a $63,312 customer acquisition cost in the publishable sample.

Those companies are far larger and run much more complicated sales processes than a typical indie SaaS, so we should not transplant their CAC into a small startup forecast. The point is the scale: B2B advertising can tolerate hundreds of dollars per lead only when each successful customer is worth a lot of money.

A $2,000 MRR startup selling $25,000 contracts could therefore have a rational reason to test LinkedIn or high-value Google searches today. A $2,000 MRR startup selling $19 subscriptions probably does not.

Should cheap consumer and AI apps wait longer before scaling ads?

Low-priced consumer and AI subscriptions should usually prove retention more carefully before scaling ads because current churn rates can wipe out the value of apparently cheap acquisition.

ChartMogul's latest retention study looked at roughly 2,700 B2B SaaS companies, 600 B2C SaaS companies and 200 AI-native companies. Among companies above $250,000 ARR, median annual net revenue retention was 82% for B2B SaaS, 49% for B2C and 48% for AI-native products.

Price makes the AI numbers even more striking.

AI-native products charging less than $50 per month had only 23% gross revenue retention and 32% net revenue retention in the study. At $50 to $249 per month, gross retention improved to 45% and NRR to 61%. Products above $250 per month reached 70% gross retention and 85% NRR.

That changes how we should think about an inexpensive $20 AI subscription acquiring customers for $30 or $40. The CAC may look fantastic on the first payment while the annual economics remain terrible.

Billing can help. ChartMogul's separate billing research found that SaaS products below $25 ARPA retained a median 62% of customers on annual plans, compared with just 41% on monthly plans.

For a cheap subscription product, moving more customers onto annual billing can therefore change how aggressively we are willing to advertise. More cash arrives upfront and more customers remain long enough for acquisition spending to pay back.

Which paid ad channel should a SaaS test first?

Google Search should usually be the first paid channel when customers already search for the problem, while Meta works better when strong creative can create interest and LinkedIn needs enough customer value to justify expensive B2B attention.

Intent is the biggest difference.

Someone typing "automated SOC 2 software" into Google is already looking for a solution. Someone scrolling Instagram probably needs to be interrupted, interested and educated before buying. LinkedIn gives us unusually precise professional targeting, but we pay heavily for access to it.

Current Meta lead campaigns average around $1.92 per click and $27.66 per lead in WordStream's latest dataset. Cheap leads sound attractive until we compare what happens after the form.

Metadata's latest B2B work makes that point particularly well. On LinkedIn, native lead forms averaged $193 per lead while campaigns sending people to an external landing page averaged $346. Document ads generated leads for about $142, compared with $200 for image ads and $265 for video.

That kind of difference is large enough to test. It still does not tell us which leads become customers.

For an early SaaS, we would choose the channel where the buying intent is easiest to explain. Search usually wins when the category already exists. Meta becomes more interesting for visual, consumer or novel products. LinkedIn starts making sense when one closed customer can comfortably absorb a four-figure experiment.

How much of your MRR should you spend on paid ads?

We would avoid setting a universal percentage of MRR for paid ads because cash tied up during CAC payback matters more than the percentage printed on the marketing budget.

Large SaaS benchmarks are especially easy to misuse here.

Benchmarkit's latest private SaaS data shows sales and marketing consuming about 47% of revenue at VC-backed companies and 33% at PE-backed companies. Those figures include salespeople, marketing salaries, software and other expenses. They tell us very little about whether a bootstrapped founder at $8,000 MRR should spend $500 or $3,000 on Meta next month.

The more useful calculation is how much acquisition cash remains unrecovered at any given time.

Suppose we spend $4,000 a month acquiring customers and those customers take six months to repay CAC. Before the system reaches steady state, roughly six months of acquisition spend, around $24,000, cycles through the payback window. Stretch payback to twelve months and that exposure approaches $48,000.

The exact cash balance changes as earlier cohorts gradually repay us, but the order of magnitude is what matters. Longer payback requires much more capital to keep growth running.

A founder with $100,000 in the bank can make a different decision from one funding ads directly from this month's Stripe balance. MRR cannot capture that difference.

How do you know paid ads are actually working?

We would call paid ads working only when the customers they produce repay acquisition cost and retain well; cheap clicks and cheap leads are secondary metrics.

A very fresh B2B benchmark shows how far advertising dashboards can drift from the business outcome.

Metadata found that $12.7 million of the 2025 spending in its dataset went into campaigns optimized around clicks or traffic. An extraordinary 99.4% of those campaigns recorded no lead at all. Only four of the 112 advertisers running them recorded even one lead from that reporting setup.

Metadata is careful about what the result means: many of those campaigns were never wired to capture or measure leads properly, so the conclusion cannot be that the $12.7 million produced zero business value. The interesting part is that huge budgets were being spent against a metric that could not answer the revenue question.

A small SaaS can make the same mistake with $500.

We want to follow each paid cohort through the whole path: click, signup, activation, payment, renewal and eventually gross profit. We should also compare paid customers with customers from organic search, referrals or direct sales. If a $70 paid customer churns twice as quickly as a $70 organic customer, their identical first-month revenue hides a very different business outcome.

Once those numbers remain stable across several cohorts, increasing the budget becomes much less speculative.

So when should you actually start doing paid ads?

For a typical bootstrapped SaaS, we would test paid ads before $5,000 MRR when there is something specific to learn, begin serious customer-acquisition tests around $5,000 to $10,000 MRR, and scale more confidently somewhere around $10,000 to $20,000+ MRR once retention and CAC payback have been proven.

If someone forces us to choose one number, $5,000 MRR is our starting threshold and $10,000 MRR is the safer one.

There are two big exceptions. A high-ticket B2B SaaS can rationally advertise at $1,000 or $2,000 MRR because one contract may cover the entire campaign. A $15 consumer or AI subscription may still have terrible paid economics at $30,000 MRR if customers leave after a few months.

The real trigger is when we can answer four questions with customer data: what does a paying customer cost us, how much gross profit does that customer generate, how long does the customer stay, and how quickly do we get our acquisition money back.

Once those answers are solid, MRR becomes much less interesting. Until then, reaching an arbitrary revenue milestone will not make paid ads safe.